Search results

[kimi-k2] [perf] support ALL2ALL communication for better performance

needs-triage

Status: Open.

#2369 In flashinfer-ai/flashinfer;

· YangXu1990uiuc opened on Jan 17, 2026

[kimi-k2] [perf] enable multi-stream in vllm

needs-triage

Status: Open.

#2368 In flashinfer-ai/flashinfer;

· YangXu1990uiuc opened on Jan 17, 2026

[kimi-k2] [perf] need to fuse small kernels

needs-triage

Status: Open.

#2367 In flashinfer-ai/flashinfer;

· YangXu1990uiuc opened on Jan 17, 2026

[feat] kernel requests for torch.compile based fusion in vLLM

needs-triage

Status: Open.

#2364 In flashinfer-ai/flashinfer;

· xinli-sw opened on Jan 16, 2026

Expand microbenchmark harness to include more APIs

benchmarkPertains to performance benchmarkingPertains to performance benchmarking

Status: Open.

#2361 In flashinfer-ai/flashinfer;

· bkryu opened on Jan 15, 2026

[Feature Request] mxfp8-blockscale (32 group size) TRTLLM Fused MoE Kernel for Blackwell (SM100)

needs-triage

Status: Open.

#2360 In flashinfer-ai/flashinfer;

· vipulSharma18 opened on Jan 15, 2026

Next fix for gemm-allreduce two-shot

cute-dsl

good first issueGood for newcomersGood for newcomers

refactor

Status: Open.

#2359 In flashinfer-ai/flashinfer;

· aleozlx opened on Jan 15, 2026

MoE unified interface

feature requestNew feature or requestNew feature or request

Status: Open.

#2358 In flashinfer-ai/flashinfer;

· aleozlx opened on Jan 15, 2026

Autotuning failsafe fallback to top1 tactic

feature requestNew feature or requestNew feature or request

Status: Open.

#2357 In flashinfer-ai/flashinfer;

· aleozlx opened on Jan 15, 2026

[bug] [trtllm_fp8_block_scale_moe] Accuracy issues with small scales

needs-triage

Status: Open.

#2356 In flashinfer-ai/flashinfer;

· jasonlizhengjian opened on Jan 15, 2026

[Feature Request] Expand Public CI GPU Architecture Coverage via GitHub Actions

needs-triage

Status: Open.

#2355 In flashinfer-ai/flashinfer;

· yongwww opened on Jan 14, 2026

[bug] Set "fa2" as the fallback backend when FP8 KV cache is enabled for Hopper GPUs

needs-triage

Status: Open.

#2348 In flashinfer-ai/flashinfer;

· nvchenghaoz opened on Jan 13, 2026